Arabidopsis thaliana wild type, var5-1 and var5-2 mutants (all in Columbia-0 background) seeds were grown on commercial soil mix (Pindstrup, Denmark) or on Murashige and Skoog (MS) (Caisson Laboratories, Smithfield, UT, United States) containing 1% sucrose, and maintained under continuous light (∼100 μmol⋅m^-2s^-1) at 22°C in controlled growth rooms. The GABI T-DNA insertion line (TAIR stock number CS323329) was obtained from the TAIR.

To map the VAR5 locus, var5-1 was crossed with Ler to generate the F2 mapping population. Bulked segregant analysis using a pool of DNAs from 96 F2 mapping plants with the var5-1 phenotype and markers on all five Arabidopsis chromosomes placed VAR5 between two molecular markers T29J13#1 and nga139 on chromosome 5. For fine mapping, 1,140 individual mapping plants were used to place the VAR5 locus into a genomic region of 92 kb. In this region, genomic sequences of putative chloroplast genes were amplified by PCR and sequences and a point mutation in At5g22830 was detected. All primers used in this study are listed in Supplementary Table S1.

The maximum quantum yield of PS II (F_V/F_M) was measured with the Open FluorCam FC800-O system (Photon Systems Instruments, Czechia) as described (Qi et al., 2016). To reveal the starch accumulation patterns in leaves, entire rosettes were boiled in 80% ethanol (v/v) until pigments were removed, then stained with fresh iodine solution (10 g/L KI and 1 g/L I₂) for 5 min, and de-stained in water for 1 h.

The middle main vein section from the third or fourth true leaves of wild type and var5-1 were hand cut and fixed in 4% (v/v) glutaraldehyde in 0.1 mM PBS buffer (pH 6.8) at 4°C overnight. After fixation, samples were dehydrated in a serial dilution of ethanol and embedded in Technovit 7100 resin (Kulzer, Wehrheim, Germany). Transverse semi-thin cross sections (3 μm) were prepared with Leica RM2265 microtome, placed on glass slides, stained with fresh iodine solution. Images of iodine stained leaf cross sections were obtained with a DM5000B microscope (Leica) equipped with a CCD camera.

Arabidopsis DNA was isolated with the CTAB method (Yu et al., 2004). Total RNA was isolated with the TRIzol reagent according to the manufacturer’s manual (Invitrogen, Carlsbad, CA, United States). cDNAs were synthesized with the Transcriptor First Strand cDNA synthesis kit (Roche, Switzerland) with random primers. For the complementation of the var5-1 mutant, full-length At5g22830 cDNAs were amplified with TaKaRa PrimeSTAR polymerase using primers 22830F and 22830R and subsequently cloned into the binary vector pBI111L to generate P_35S: At5g22830 (Yu et al., 2004). This construct was transformed in the var5-1 mutant using the floral dip method (Clough and Bent, 1998), and transgenic plants were screened on 1/2 MS plates containing 50 mg/L kanamycin.

AtMRS2-11/AtMGT10/VAR5 homologous proteins in different species were identified using the BLASTP program of the National Center for Biotechnology Information. Multiple sequence alignment and phylogenetic analysis were performed using the MEGA6 software (Tamura et al., 2013). Gene structures were constructed based on gene sequences retrieved from the Phytozome resource¹ (Goodstein et al., 2012).

To construct a transient expression vector P_35S:VAR5-GFP, full-length coding sequences of AtMRS2-11/AtMGT10/VAR5 were amplified with primers 22830GFPF and 22830GFPR and cloned into pUC18-GFP vector with the GFP coding sequences fused at its 3′ terminus, and the expression of VAR5-GFP was under the CaMV 35S promoter. The 35S promoter in P_35S:VAR5-GFP was replaced by a ∼1.9 kb promoter region of VAR5 to generate P_{V AR5}:VAR5-GFP. These constructs, as well as the P_35S:GFP control vector, were transformed into protoplasts isolated from 4-week-old wild-type plants as described (Yoo et al., 2007). Chlorophyll fluorescence and GFP signals were examined by a Nikon A1 confocal microscope (Nikon, Japan).

Intact chloroplasts were isolated from 4-week-old Arabidopsis plants that were dark treated for 16 h (Kunst, 1998). To fractionate sub-chloroplast compartments, intact chloroplasts were broken by passing through a 24-gauge syringe in hypotonic buffer (10 mM Hepes-KOH pH7.6, 5 mM MgCl₂). The lysate was separated on sucrose gradients by ultracentrifugation with SW41 Ti rotor at 58,000 × g for 2 h (Beckman Coulter L100-XP), and fractions were collected as described (Qi et al., 2016). For BN-PAGE, chloroplast samples equivalent to 10 μg of total chlorophylls were solubilized in 25BTH20G buffer [25 mM Bis-Tris-HCl, pH 7.0, 20% (w/v) glycerol] containing 1.0% (w/v) of n-dodecyl-β-D-maltoside (β-DM) (Järvi et al., 2011). The solubilized chloroplast samples were resolved on 3–8% BN-PAGE with a constant voltage of 70 V at 4°C, and the first dimension gel strips were denatured and resolved on 10% SDS-PAGE. For immunoblot analysis, proteins separated by SDS-PAGE were transferred onto 0.2 μm PVDF membranes with the Trans-blot Semi-Dry Transfer Cell (Bio-Rad) following the manufacturer’s manual. PVDF membranes were incubated with specific primary antibodies (Agrisera), followed by horseradish peroxidase-conjugated secondary antibodies. Signals were detected with chemiluminescence solutions and ChemiDoc Imaging System (Bio-Rad).

During the characterization of an Arabidopsis SALK T-DNA insertion line, one unique reticulate (‘net-like’) leaf variegation mutant was recovered (Figure 1A; early stage of this work was carried out in the lab of Dr. Steve Rodermel at Iowa State University). A second and similar reticulate leaf variegation mutant was identified in our genetic screens for chloroplast development mutants (Supplementary Figures S1A–D, Qi et al., 2016). Genetic tests revealed that these two mutants were allelic and we thus designated the two mutants as variegated5-1 (var5-1) and variegated5-2 (var5-2), respectively, following the variegation mutant naming sequence (Martínez-Zapater, 1993; Yu et al., 2007) (Supplementary Figures S1G,H). Both mutant alleles showed a similar reticulate leaf variegation phenotype (Figures 1A,B and Supplementary Figure S1A). This unique leaf variegation phenotype is more pronounced in adult leaves than juvenile leaves, as the seventh true leaf of var5-1 showed a clear reticulate phenotype, with overall yellow paraveinal regions and green interveinal regions (Figure 1B). Taken together, var5-1 and var5-2 represent two allelic mutants that display a unique leaf reticulation phenotype.

The reticulate leaf color phenotype suggests that photosynthetic capacities are compromised in var5-1 mutants. To correlate the visible leaf color phenotype with the photosynthetic status of the mutant, we measured parameters that indicate photosynthetic performance. For the light reactions, we measured the maximum quantum yield of PS II, also known as F_V/F_M, using a whole plant chlorophyll fluorescence imaging system. The average F_V/F_M of 15 min dark adapted wild-type whole plants was 0.82, while that of var5-1 was 0.66, suggesting that photosynthesis is compromised in the mutant (Figure 1C). In addition, while a single leaf of wild-type plant showed a uniform distribution of F_V/F_M values, the F_V/F_M from var5-1 also showed a ‘net-like’ pattern, with paraveinal regions showing higher value while interveinal regions having lower F_V/F_M (Figure 1D). We next examined the accumulation of starch in the wild type and var5-1 plants as one indicator of photosynthetic productivities. Interestingly, the starch accumulation in the var5-1 mutant shows clearly a reticulate pattern, clearly matching both the leaf reticulation and F_V/F_M patterns (Figures 1E,F). In addition, iodine stained leaf cross sections showed that chloroplasts in the palisade mesophyll cells on the top of vascular bundles in the var5-1 mutant, were significantly less stained compared to that of wild type (Supplementary Figures S2A–D), suggesting the development of chloroplasts in the yellow paraveinal regions are compromised in the var5-1 mutant. These data indicate that the mutation in var5-1 leads to a unique expression of leaf reticulation and a pronounced reduction of photosynthetic capacities as indicated by PSII activities and starch accumulation in paraveinal regions than in interveinal regions.

Genetic analysis revealed that the reticulate phenotype of var5-1 was likely caused by a single nuclear recessive mutation, and map-based cloning procedures were used to clone the VAR5 locus (Figure 2). Initial mapping with molecular markers on all five chromosomes and subsequent fine mapping using 1,140 mapping plants placed VAR5 in a ∼92 kb genomic region between markers MDJ22#2 and MRN17#1 on chromosome 5 (Figure 2A). Based on the leaf color phenotype, genomic DNA sequences of nuclear genes for chloroplast proteins in this interval were examined. A single G to A transition mutation was identified in the first nucleotide of intron 9 in At5g22830 (Figure 2A). At the mRNA level, amplification products spanning the mutation site appeared longer in the mutant than in wild type, suggesting pre-mRNA splicing might be affected in var5-1 (Figure 2B). Sequencing of the full-length At5g22830 cDNA in var5-1 revealed the nature of pre-mRNA splicing defects in var5-1. The mutation in the intron 9 led to the retention of the entire intron 9 in the At5g22830 transcript (Figure 2C). Surprisingly, a partial segment of intron 6 (the first 17 bp of the 328 bp wild-type intron 6) was also retained (Figure 2C). The translation of this aberrant mRNA would in theory give rise to a protein with the first 262 amino acids of the annotated 459 amino acids of wild-type protein plus four additional amino acids before a premature stop codon (Figure 2D). In theory, this early stop codon would abolish the C-terminal transmembrane domains and the conserved GMN motif (Figure 2E). We also determined the mutation site in var5-2 and identified a G to A mutation at the last nucleotide of intron 4, which also caused abnormal At5g22830 transcript in var5-2 (Supplementary Figures S1E,F). This mutation led to the retention of the entire intron 4 (80 bp) in the At5g22830 transcript and a putative translation product with the first 208 amino acids of wild-type protein before a premature stop codon (Supplementary Figures S1E,F).

To confirm that the reticulate phenotype of var5-1 was indeed caused by the mutation identified in At5g22830, a binary vector (P_35S: At5g22830) containing the wild-type At5g22830 cDNA under the control of the CaMV 35S promoter was constructed and transformed into var5-1. Multiple transgenic lines that showed completely green leaf color and the presence of wild-type cDNA amplification products were identified, indicating that the reticulate phenotype of var5-1 can be rescued by the wild-type At5g22830 (Figures 3A–D). At the mRNA level, the complementation lines showed the presence of wild-type At5g22830 cDNA (Figure 3E). At the protein level, we generated polyclonal antibodies against amino acid residues from 61 to 263 of At5g22830 and Western blotting analyses detected the accumulation of At5g22830 protein in the wild type, but not in the var5-1 mutant (Figure 3F). In contrast, At5g22830 protein accumulation was greatly enhanced in the var5-1 P_35S:At5g22830 complementation line (Figure 3F). Similar results were obtained with var5-2 mutant (Supplementary Figure S1G). The identifications of mutations sites in var5-1 and var5-2, the molecular complementation and the western data show that At5g22830 is VAR5.

VAR5/At5g22830 was previously reported to encode a putative CorA family Mg²⁺ transporter that is also known as AtMRS2-11/AtMGT10 (Schock et al., 2000; Li et al., 2001; Drummond et al., 2006). At the protein level, AtMRS2-11/AtMGT10/VAR5 contains the characteristic features of CorA Mg²⁺ transporters, including the two C-terminal transmembrane domains and in between the Mg²⁺-binding GMN motif (Lunin et al., 2006; Pfoh et al., 2012). There are at least 10 members of CorA family Mg²⁺ transporters in Arabidopsis thaliana, and some of them are predicted to be targeted into organelles including the chloroplast and the mitochondrion (Gebert et al., 2009). We analyzed the relationship between AtMRS2-11/AtMGT10/VAR5 and its putative chloroplast-localized homologs from various eukaryotic photosynthetic organisms, as well as CorA homologs from bacteria (Figure 4A). Overall, AtMRS2-11/AtMGT10/VAR5 homologs from different organisms are highly conserved (Figure 4A). In addition, these proteins can be grouped into major clades including eudicot, monocot, algae and prokaryotic organisms (Figure 4A). The evolutionary relationship between these genes were also obvious at the DNA level as putative higher plant AtMRS2-11/AtMGT10/VAR5 related genes shared identical gene structures, suggesting these genes emerged before the divergence of eudicot and monocot plants (Figure 4B). In contrast, AtMRS2-11/AtMGT10/VAR5 homologs from green algae did not share conserved gene structures with their plant counterparts (Figure 4B). These data suggest that the CorA family Mg²⁺ transporter is prevalent and highly conserved, consistent with the critical roles Mg²⁺ plays in prokaryotic and eukaryotic cells.

AtMRS2-11/AtMGT10/VAR5 is predicted by the ChloroP program to contain a putative chloroplast transit peptide (Emanuelsson et al., 1999). Experimentally, it was shown to be localized on the periphery of chloroplasts (Drummond et al., 2006). We took two approaches to independently confirm the sub-cellular localization of AtMRS2-11/AtMGT10/VAR5. First, we used the protoplast transient expression assay to examine the localization of VAR5-GFP under the control of the 35S promoter (P_35S:VAR5-GFP) and the endogenous VAR5 promoter (P_{V AR5}:VAR5-GFP), respectively (Figure 5A). Control vector (P_35S:GFP) gave GFP signals mainly in the cytosol and the nucleus (Figure 5A). In contrast, both P_35S:VAR5-GFP and P_{V AR5}:VAR5-GFP gave GFP signals that outlined the periphery of the chloroplast, as well as some puncta inside the chloroplast, suggesting both promoters can direct VAR5 to sub-chloroplast locations that likely correspond to the chloroplast envelope (Figure 5A). Next, we determined the AtMRS2-11/AtMGT10/VAR5 sub-chloroplast localization biochemically (Figure 5B). Intact chloroplasts were fractionated into the thylakoids (Thy), the stroma (Str), the envelope fraction 1 (Env1) and 2 (Env2) with sucrose gradient centrifugations (Qi et al., 2016). Using antibodies against chloroplast proteins with known sub-chloroplast localizations, we showed that RbcL (Rubisco large subunit) was enriched in the stroma fraction, and Lhcb (PSII light harvesting complex protein) was enriched in the thylakoids fraction as expected. The enrichment of chloroplast envelope membranes in the Env2 fraction was indicated by chloroplast envelope proteins Tic40 and Toc34 (components of the Translocon at the Inner/Outer Envelope Membrane of Chloroplasts, TIC/TOC) (Figure 5B; Nakai, 2015). Using VAR5 antibodies that we had generated, we found that the majority of AtMRS2-11/AtMGT10/VAR5 was present in the Env2 fraction, similar to the distribution patterns of TIC/TOC components, suggesting that it is likely a chloroplast envelope membrane protein (Figure 5B).

The crystal structure of the thermophilic bacterium Thermotoga maritima CorA Mg²⁺ transporter indicates that it forms a homopentamer (Lunin et al., 2006). To determine whether AtMRS2-11/AtMGT10/VAR5 forms complexes in Arabidopsis, two-dimensional BN-PAGE was carried out (Järvi et al., 2011). BN-PAGE can effectively separate chloroplast membrane protein complexes, and the estimated sizes of the major photosynthetic protein complexes are known (Figure 6). For instance, the sizes of the four large PSII supercomplexes were estimated to be 880–1,400 kDa (Caffarri et al., 2009; Hou et al., 2015), while the PSI monomer and the rubisco holoenzyme migrate at approximately 600 and 540 kDa, respectively (Berghöfer and Klösgen, 1999; Järvi et al., 2011). Western blot analysis with second dimension gels using antibodies against AtpA (the α subunit of the chloroplast ATP synthase), Lhcb, PetC (a subunit of the cytochrome b₆f complex) and PsaF (a subunit of the PSI) clearly identified various photosynthetic complexes at expected sizes (Figure 6). Interestingly, AtMRS2-11/AtMGT10/VAR5 antibodies detected a streak of signals in the second dimension gels (Figure 6). Based on our results, it is clear that different forms of complexes containing AtMRS2-11/AtMGT10/VAR5 exist in the chloroplast, and the sizes of these complexes are clearly larger than the PSI monomer (∼600 kDa) (Figure 6). The distribution of AtMRS2-11/AtMGT10/VAR5 complexes overlapped partially with the PSII supercomplexes, which suggest the complexes could reach over 800 kDa (Figure 6). This is in stark contrast to the CorA complexes in Thermotoga maritime, which have molecular weights around 210 kDa as homopentamers (Lunin et al., 2006; Matthies et al., 2016). We did not detect AtMRS2-11/AtMGT10/VAR5 signals in var5-1 plants, agreeing with our one dimension SDS-PAGE results (Figure 2). These data suggest that AtMRS2-11/AtMGT10/VAR5 forms large complexes and the complex compositions are different in Arabidopsis from those in bacteria.